Rufus became Alexa for Shopping
Amazon retired Rufus as a standalone assistant and launched Alexa for Shopping in its place on May 13, 2026, available to all US customers in the Amazon app and on amazon.com. It's more than a name change: the new assistant sits right in the main search bar, generates AI summaries above the results and runs side-by-side product comparisons inside the results page itself — more integrated into the shopping journey than the original Rufus was.
What it reads to answer shoppers
Alexa for Shopping uses a language model to interpret the shopper's intent and generate its own search from that context — unlike traditional keyword search. To answer questions about a specific product, it reads the title, bullet points, A+ Content, images and customer reviews, plus other data Amazon already has about the product.
How to optimize for that reading
Products with complete structured attributes (material, intended use, certifications) tend to perform better than listings built on keyword stuffing. Copy aligned with shopping intent — "great for travel," "quiet enough for apartments," "gentle on sensitive skin" — converts better than generic adjectives. In practice: if the title, bullets, A+ and images make it clear what the product is and who it's for, the AI has cleaner material to summarize — your content should answer the questions a shopper would ask out loud, not just the ones they'd type into search.
What COSMO is
COSMO is far less known to the public and never had a launch announcement like Rufus — the available information comes from a research paper Amazon itself published at SIGMOD 2024. It's a system that uses language models to mine "common-sense" knowledge from real customer search and purchase behavior, building a graph of relationships like "slip-resistant shoes → suitable for → pregnant women" or "camera case + screen protector → used for → protecting the camera" — even when those associations were never written literally in any listing.
According to the paper, COSMO has been tested on about 10% of US search traffic across 18 major product categories, with a 60% improvement in search relevance (macro F1) and a 0.7% sales lift in the A/B test. It's important to understand that it does not replace the A9 search algorithm — it's an additional intelligence layer that complements it.
What it changes in practice
Because COSMO infers use and audience from shopping behavior (not just listing text), a product can start showing up for implicit-use searches that were never a direct target of keyword optimization — as long as real purchase history supports the association. That doesn't mean dropping traditional keyword optimization: it means structured, accurate and consistent information (attributes, the right category, genuine reviews about how the product is used) helps both Alexa for Shopping's direct reading and COSMO's behavioral inference.
How EcomVision helps
EcomVision's Listing Optimizer structures your title, bullets and A+ with the attributes and intent-driven language that both Alexa for Shopping and COSMO use to understand a product — without relying on keyword stuffing.
